Schema Alignment Skill

Detect drift between database schemas and code data models. This skill identifies missing columns, type mismatches, orphaned migrations, and naming inconsistencies.

Design Principle

This skill is framework-generic. It works with any ORM or database:

• SQLAlchemy (Python)
• Django ORM (Python)
• Prisma (TypeScript/JavaScript)
• TypeORM (TypeScript)
• Drizzle (TypeScript)
• Alembic migrations
• Prisma migrations
• Django migrations

Variables

Variable	Default	Description
SCHEMA_SOURCE	auto	Schema source: auto, migrations, live_db, models
SEVERITY_THRESHOLD	medium	Report issues at this level or higher
AUTO_FIX	false	Attempt to generate fix suggestions
INCLUDE_TYPES	true	Include type mismatch detection

Instructions

MANDATORY - Follow the Workflow steps below in order.

1. Detect database technology and ORM in use
2. Extract schema from migrations or live database
3. Extract data models from code
4. Compare and identify drift
5. Generate alignment report

Red Flags - STOP and Reconsider

If you're about to:

• Modify the database schema directly without a migration
• Assume a column exists without checking the schema
• Skip type checking because "it works in tests"
• Ignore nullable/not-null mismatches

STOP -> Check schema alignment -> Generate migration if needed -> Then proceed

Workflow

1. Detect Stack

Identify the database and ORM:

Check for these indicators:

| File/Dependency | Technology |
|-----------------|------------|
| alembic.ini, alembic/ | Alembic (SQLAlchemy) |
| prisma/schema.prisma | Prisma |
| manage.py + migrations/ | Django |
| ormconfig.json | TypeORM |
| drizzle.config.ts | Drizzle |
| supabase/migrations/ | Supabase (PostgreSQL) |

2. Extract Database Schema

Option A: From Migrations (Preferred)

Parse migration files to reconstruct current schema:

# Alembic example
from alembic.script import ScriptDirectory
from alembic.config import Config

config = Config("alembic.ini")
scripts = ScriptDirectory.from_config(config)
# Walk revisions to build schema

Option B: From Live Database

Query information_schema (if accessible):

SELECT table_name, column_name, data_type, is_nullable
FROM information_schema.columns
WHERE table_schema = 'public';

Option C: From Model Definitions

Parse ORM model files directly.

3. Extract Code Models

Parse model definitions from code:

SQLAlchemy

# Look for patterns like:
class User(Base):
    __tablename__ = 'users'
    id = Column(Integer, primary_key=True)
    email = Column(String(255), nullable=False)

Prisma

model User {
  id    Int     @id @default(autoincrement())
  email String  @unique
  name  String?
}

Pydantic/TypeScript Types

Also extract related types:

• Pydantic models
• TypeScript interfaces
• BAML type definitions

4. Compare and Detect Drift

Run comparisons:

Check	Source A	Source B	Issue Type
Missing column	DB schema	ORM model	MISSING_IN_MODEL
Missing column	ORM model	DB schema	MISSING_IN_DB
Type mismatch	DB type	Code type	TYPE_MISMATCH
Nullable mismatch	DB nullable	Model nullable	NULLABLE_MISMATCH
Name mismatch	snake_case	camelCase	NAMING_DRIFT
Missing migration	Model change	Migration files	MISSING_MIGRATION
FK constraint	DB constraint	ORM relationship	FK_MISMATCH

5. Generate Report

Output format:

# Schema Alignment Report

**Generated**: 2025-12-24T10:00:00Z
**Database**: PostgreSQL (via Supabase)
**ORM**: SQLAlchemy 2.0

## Summary

| Severity | Count |
|----------|-------|
| HIGH | 2 |
| MEDIUM | 3 |
| LOW | 5 |

## Issues

### 1. MISSING_IN_MODEL (HIGH)

**Table**: `curation_jobs`
**Column**: `retry_count` (INTEGER NOT NULL DEFAULT 0)
**Model**: `src/models/curation_job.py:CurationJob`

The column exists in the database but is not defined in the ORM model.

**Fix**:
```python
retry_count: Mapped[int] = mapped_column(Integer, default=0)

2. TYPE_MISMATCH (MEDIUM)

Table: books Column: isbn (VARCHAR(13)) Model: src/models/book.py:Book.isbn -> str

Database constrains to 13 characters but model allows unbounded string.

Fix:

isbn: Mapped[str] = mapped_column(String(13))

3. MISSING_MIGRATION (LOW)

Model Change: User.preferences added (JSONB) Migration: Not found

A new column was added to the model but no migration exists.

Fix:

alembic revision --autogenerate -m "add user preferences"

## Cookbook

### SQLAlchemy Detection
- IF: Parsing SQLAlchemy models
- THEN: Read and execute `./cookbook/sqlalchemy-detection.md`

### Prisma Detection
- IF: Parsing Prisma schema
- THEN: Read and execute `./cookbook/prisma-detection.md`

### Alembic Migrations
- IF: Generating migration fix
- THEN: Read and execute `./cookbook/alembic-migration.md`

## Issue Severity Matrix

| Issue Type | Default Severity | Upgrade If |
|------------|-----------------|------------|
| MISSING_IN_MODEL | HIGH | Column is NOT NULL |
| MISSING_IN_DB | MEDIUM | Model references it |
| TYPE_MISMATCH | MEDIUM | Could cause data loss |
| NULLABLE_MISMATCH | LOW | NOT NULL in code, nullable in DB |
| NAMING_DRIFT | LOW | - |
| MISSING_MIGRATION | LOW | - |
| FK_MISMATCH | MEDIUM | Causes ORM errors |

## Integration

### With /ai-dev-kit:check-schema

Direct invocation:

```bash
# Full check
/ai-dev-kit:check-schema

# Check specific tables
/ai-dev-kit:check-schema --tables=users,orders

# Generate fixes
/ai-dev-kit:check-schema --auto-fix

# Output to file
/ai-dev-kit:check-schema --output=alignment-report.md

With /ai-dev-kit:execute-lane

Runs as pre-flight check for database-related lanes:

Lane: SL-DB (Database Schema)

Pre-flight checks:
1. ✓ Git worktree clean
2. ✗ Schema alignment check failed
   - 2 HIGH severity issues found
   - See alignment-report.md

Action: Resolve schema issues before proceeding.

With /ai-dev-kit:plan-phase

Runs during phase planning:

Planning Phase P1...

Schema Alignment: ⚠️ 3 issues detected
- 1 missing migration
- 2 type mismatches

Recommendation: Add schema alignment task to SL-DB lane.

Output Schema

{
  "$schema": "http://json-schema.org/draft-07/schema#",
  "type": "object",
  "properties": {
    "generated_at": {"type": "string", "format": "date-time"},
    "database": {"type": "string"},
    "orm": {"type": "string"},
    "summary": {
      "type": "object",
      "properties": {
        "high": {"type": "integer"},
        "medium": {"type": "integer"},
        "low": {"type": "integer"}
      }
    },
    "issues": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "type": {"type": "string"},
          "severity": {"enum": ["HIGH", "MEDIUM", "LOW"]},
          "table": {"type": "string"},
          "column": {"type": "string"},
          "model_location": {"type": "string"},
          "description": {"type": "string"},
          "fix": {"type": "string"}
        }
      }
    }
  }
}

Best Practices

1. Run regularly: Check schema alignment before each PR
2. CI integration: Add to CI pipeline for automatic detection
3. Migration hygiene: Always generate migrations for model changes
4. Type consistency: Use explicit types in models matching DB constraints
5. Document drift: If drift is intentional, document why